pith. sign in

arxiv: 2604.17186 · v1 · submitted 2026-04-19 · 💻 cs.SE · cs.AI· cs.ET· cs.HC· cs.MA

Persona-Based Requirements Engineering for Explainable Multi-Agent Educational Systems: A Scenario Simulator for Clinical Reasoning Training

Weibing Zheng , Laurah Turner , Jess Kropczynski , Matthew Kelleher , Murat Ozer , Shane Halse This is my paper

Pith reviewed 2026-05-10 06:31 UTC · model grok-4.3

classification 💻 cs.SE cs.AIcs.ETcs.HCcs.MA
keywords persona-based requirements engineeringexplainable multi-agent systemsclinical reasoning trainingmulti-agent educational systemshuman-centered AImedical educationscenario simulatorAI explainability
0
0 comments X

The pith

Persona-based requirements engineering ensures multi-agent educational systems for clinical training are explainable and aligned with real scenarios from the start.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The paper proposes a human-first framework that weaves personas and user stories into every step of requirements engineering for multi-agent educational systems. This captures the distinct needs of medical students, educators, and AI agents such as the patient simulator and diagnostic, exam, and evaluation agents. The personas shape the agents' goals, knowledge bases, and interaction rules, which in turn define the explainability features required for trust and transparency. In a concrete clinical reasoning simulator, a post-usage survey showed more than 78 percent of medical students reporting improved skills. The work therefore claims that starting with personas links technical specifications to non-technical users early enough to produce systems that feel authentic and interpretable.

Core claim

The central claim is that a persona-driven explainable MAES requirements engineering framework, demonstrated on a clinical reasoning training simulator, integrates personas and user stories for medical educators, students, an AI patient agent, and clinical agents to define goals, underlying models, knowledge bases, and interactions; this process directly informs explainability requirements that produce systems aligned with authentic clinical scenarios, as indicated by over 78 percent of students reporting skill improvement.

What carries the argument

The persona-driven explainable MAES RE framework, which uses distinct personas for each stakeholder and agent type to capture goals and behaviors that determine agent interactions and explainability requirements.

If this is right

  • Explainability requirements are identified and incorporated at the requirements stage rather than retrofitted after implementation.
  • Agent behaviors and knowledge bases are constrained by real stakeholder goals, producing interactions that match clinical practice.
  • Non-technical users such as medical students can directly influence system design, increasing perceived trustworthiness.
  • The resulting MAES can be partially open-sourced for further validation and reuse in similar training contexts.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • The same persona integration step could be tested in other multi-agent domains such as legal reasoning or engineering design education to check whether early alignment improves user-reported outcomes.
  • Developers adopting this framework might observe higher long-term retention of AI tools in medical curricula compared with systems engineered without explicit persona input.
  • A follow-up study could measure whether the explainability features derived from personas reduce the frequency of students misunderstanding agent recommendations during simulated cases.

Load-bearing premise

That integrating personas and user stories throughout the requirements engineering process will consistently produce explainable and effective multi-agent educational systems.

What would settle it

A controlled comparison developing an otherwise identical clinical reasoning MAES without personas and measuring whether the percentage of students reporting skill improvement and perceived explainability drops significantly below 78 percent.

Figures

Figures reproduced from arXiv: 2604.17186 by Jess Kropczynski, Laurah Turner, Matthew Kelleher, Murat Ozer, Shane Halse, Weibing Zheng.

Figure 1
Figure 1. Figure 1: Persona-Based RE Framework for Explainable MAES [PITH_FULL_IMAGE:figures/full_fig_p003_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: AI Agents and their Relationships in MAES for CSS [PITH_FULL_IMAGE:figures/full_fig_p004_2.png] view at source ↗
read the original abstract

As Artificial Intelligence (AI) and Agentic AI become increasingly integrated across sectors such as education and healthcare, it is critical to ensure that Multi-Agent Education System (MAES) is explainable from the early stages of requirements engineering (RE) within the AI software development lifecycle. Explainability is essential to build trust, promote transparency, and enable effective human-AI collaboration. Although personas are well-established in human-computer interaction to represent users and capture their needs and behaviors, their role in RE for explainable MAES remains underexplored. This paper proposes a human-first, persona-driven, explainable MAES RE framework and demonstrates the framework through a MAES for clinical reasoning training. The framework integrates personas and user stories throughout the RE process to capture the needs, goals, and interactions of various stakeholders, including medical educators, medical students, AI patient agent, and clinical agents (physical exam agent, diagnostic agent, clinical intervention agent, supervisor agent, evaluation agent). The goals, underlying models, and knowledge base shape agent interactions and inform explainability requirements that guided the clinical reasoning training of medical students. A post-usage survey found that more than 78\% of medical students reported that MAES improved their clinical reasoning skills. These findings demonstrate that RE based on persona effectively connects technical requirements with non-technical medical students from a human-centered approach, ensuring that explainable MAES are trustworthy, interpretable, and aligned with authentic clinical scenarios from the early stages of the AI system engineering. The partial MAES for the clinical scenario simulator is~\href{https://github.com/2sigmaEdTech/MAS/}{open sourced here}.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

2 major / 2 minor

Summary. The manuscript proposes a persona-driven requirements engineering (RE) framework for explainable multi-agent educational systems (MAES) in clinical reasoning training. It integrates personas for stakeholders (medical educators, students, AI patient agent, diagnostic agent, clinical intervention agent, supervisor agent, evaluation agent) and user stories to capture needs, goals, and interactions, which then inform explainability requirements and agent behaviors. A partial MAES scenario simulator is demonstrated and partially open-sourced; a post-usage survey reports that more than 78% of medical students indicated improved clinical reasoning skills, supporting the claim that persona-based RE produces trustworthy, interpretable MAES aligned with authentic clinical scenarios from the early stages of development.

Significance. If the empirical support is strengthened, the work could usefully advance human-centered RE practices for agentic AI systems in education and healthcare by showing how early persona integration can embed explainability and stakeholder alignment. The open-sourcing of the partial implementation is a concrete strength that aids reproducibility and extension.

major comments (2)
  1. [Evaluation / results section (and abstract)] The central empirical claim rests on a post-usage survey reporting >78% self-reported improvement in clinical reasoning skills. No sample size, demographics, validated instrument, statistical tests, control arm, baseline measures, or objective performance indicators (e.g., diagnostic accuracy or script concordance scores) are provided. This leaves the causal attribution to the persona-based RE framework unsupported and vulnerable to confounds such as novelty or selection bias.
  2. [Framework and case-study sections] The framework description does not supply explicit traceability from individual persona attributes or user stories to the concrete explainability requirements and agent interaction rules (e.g., how the 'AI patient agent' persona directly shapes the diagnostic agent's output format or transparency features). Without this mapping, the assertion that the approach 'connects technical requirements with non-technical medical students' and ensures alignment with clinical scenarios remains descriptive rather than demonstrated.
minor comments (2)
  1. [Abstract] The GitHub link is given in the abstract but lacks a permanent identifier or version tag; adding a citation or archive link would improve long-term accessibility.
  2. [Introduction] Terminology such as 'Agentic AI' and 'MAES' would benefit from explicit definitions or references on first use to aid readers outside the immediate subfield.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for their constructive and detailed feedback. We address each major comment below and outline the revisions we will make to improve the manuscript.

read point-by-point responses

  1. Referee: [Evaluation / results section (and abstract)] The central empirical claim rests on a post-usage survey reporting >78% self-reported improvement in clinical reasoning skills. No sample size, demographics, validated instrument, statistical tests, control arm, baseline measures, or objective performance indicators (e.g., diagnostic accuracy or script concordance scores) are provided. This leaves the causal attribution to the persona-based RE framework unsupported and vulnerable to confounds such as novelty or selection bias.

    Authors: We agree that the survey description is limited and does not support causal claims about the persona-based RE framework. The evaluation was exploratory and relied on self-reported feedback without the methodological details noted. In the revised manuscript we will expand the results section and abstract to characterize the survey explicitly as preliminary, note the lack of controls, objective measures, and statistical tests, and qualify the 78% figure to avoid implying direct attribution or generalizability. These changes will mitigate the risk of overstatement while preserving the reported user perception as supporting evidence for further study. revision: yes

  2. Referee: [Framework and case-study sections] The framework description does not supply explicit traceability from individual persona attributes or user stories to the concrete explainability requirements and agent interaction rules (e.g., how the 'AI patient agent' persona directly shapes the diagnostic agent's output format or transparency features). Without this mapping, the assertion that the approach 'connects technical requirements with non-technical medical students' and ensures alignment with clinical scenarios remains descriptive rather than demonstrated.

    Authors: We accept that the current text lacks explicit traceability and therefore remains largely descriptive. We will add a dedicated mapping subsection (or table) in the framework description that links specific persona attributes and user stories to the resulting explainability requirements and agent behaviors. For example, we will show how attributes of the AI patient agent (goals, knowledge base) directly inform the diagnostic agent's output format and transparency mechanisms, thereby demonstrating the connection between stakeholder needs and technical implementation. revision: yes

Circularity Check

0 steps flagged

No circularity: framework and survey are independent descriptions with no derivations or self-referential reductions.

full rationale

The paper proposes a persona-driven RE framework for explainable MAES, illustrates it via a clinical reasoning simulator with defined agent roles, and reports a post-usage survey result (>78% self-reported improvement). No equations, fitted parameters, predictions, or uniqueness theorems appear. The central claim rests on the framework description plus the survey data; neither reduces to the other by construction, nor does any step invoke self-citation chains or ansatzes that collapse the argument. This is a standard descriptive/empirical paper whose derivation chain is self-contained against external benchmarks.

Axiom & Free-Parameter Ledger

0 free parameters · 0 axioms · 0 invented entities

This is a conceptual framework and evaluation paper with no formal mathematics, fitted parameters, or postulated entities. All elements are described in prose without quantitative models.

pith-pipeline@v0.9.0 · 5630 in / 1109 out tokens · 36155 ms · 2026-05-10T06:31:30.686542+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

30 extracted references · 30 canonical work pages

  1. [1]

    A Fuzzy Supervisor Agent Design for Clinical Reasoning Assistance in a Multi-Agent Educational Clinical Scenario Simulation,

    W. Zheng, L. Turner, J. Kropczynski, M. Ozer, S. Overla, and S. Halse, “A Fuzzy Supervisor Agent Design for Clinical Reasoning Assistance in a Multi-Agent Educational Clinical Scenario Simulation,” Jul. 2025. [Online]. Available: http://arxiv.org/abs/2507.05275

    work page arXiv 2025

  2. [2]

    The User-Centered Iterative Design of an LLM-Powered Educational Scenario Simulator for Clinical Reasoning,

    W. Zheng, L. Turner, J. Kropczynski, M. Ozer, M. Kelleher, S. Overla, and S. Halse, “The User-Centered Iterative Design of an LLM-Powered Educational Scenario Simulator for Clinical Reasoning,”CS & IT Conference Proceedings, vol. 15, no. 10, May 2025. [Online]. Available: http://csitcp.com/abstract/15/1510csit19

    work page 2025

  3. [3]

    Interpreting Black-Box Models: A Review on Explainable Artificial Intelligence,

    V . Hassija, V . Chamola, A. Mahapatra, A. Singal, D. Goel, K. Huang, S. Scardapane, I. Spinelli, M. Mahmud, and A. Hussain, “Interpreting Black-Box Models: A Review on Explainable Artificial Intelligence,” Cognitive Computation, vol. 16, no. 1, pp. 45–74, Jan. 2024

    work page 2024

  4. [4]

    How mature is requirements engineering for AI-based systems? A systematic mapping study on practices, challenges, and future research directions,

    U.-e. Habiba, M. Haug, J. Bogner, and S. Wagner, “How mature is requirements engineering for AI-based systems? A systematic mapping study on practices, challenges, and future research directions,”Require- ments Engineering, vol. 29, no. 4, pp. 567–600, Dec. 2024

    work page 2024

  5. [5]

    Can Requirements Engi- neering Support Explainable Artificial Intelligence? Towards a User- Centric Approach for Explainability Requirements,

    U.-E. Habiba, J. Bogner, and S. Wagner, “Can Requirements Engi- neering Support Explainable Artificial Intelligence? Towards a User- Centric Approach for Explainability Requirements,” in2022 IEEE 30th International Requirements Engineering Conference Workshops (REW), Aug. 2022, pp. 162–165

    work page 2022

  6. [6]

    Scenario- Based Requirements Elicitation for User-Centric Explainable AI,

    D. Cirqueira, D. Nedbal, M. Helfert, and M. Bezbradica, “Scenario- Based Requirements Elicitation for User-Centric Explainable AI,” in Machine Learning and Knowledge Extraction, A. Holzinger, P. Kiese- berg, A. M. Tjoa, and E. Weippl, Eds. Cham: Springer International Publishing, 2020, pp. 321–341

    work page 2020

  7. [7]

    ChatGPT for Learning HCI Tech- niques: A Case Study on Interviews for Personas,

    J. Barambones, C. Moral, A. de Antonio, R. Imbert, L. Mart ´ınez- Normand, and E. Villalba-Mora, “ChatGPT for Learning HCI Tech- niques: A Case Study on Interviews for Personas,”IEEE Transactions on Learning Technologies, vol. 17, pp. 1460–1475, 2024

    work page 2024

  8. [8]

    Russell and P

    S. Russell and P. Norvig,Artificial Intelligence: A Modern Approach, 4th ed. Hoboken, NJ: Pearson, 2021

    work page 2021

  9. [9]

    Llm-powered autonomous agents,

    L. Weng, “Llm-powered autonomous agents,”lilianweng.github.io, Jun 2023. [Online]. Available: https://lilianweng.github.io/posts/ 2023-06-23-agent/

    work page 2023

  10. [10]

    Large Language Model based Multi-Agents: A Survey of Progress and Challenges,

    T. Guo, X. Chen, Y . Wang, R. Chang, S. Pei, N. V . Chawla, O. Wiest, and X. Zhang, “Large Language Model based Multi-Agents: A Survey of Progress and Challenges,” Apr. 2024

    work page 2024

  11. [11]

    A grounded interaction protocol for explainable artificial intelligence,

    P. Madumal, T. Miller, L. Sonenberg, and F. Vetere, “A grounded interaction protocol for explainable artificial intelligence,”arXiv preprint arXiv:1903.02409, 2019

    work page arXiv 1903

  12. [12]

    Ai agent for education: von neumann multi-agent system framework,

    Y .-H. Jiang, R. Li, Y . Zhou, C. Qi, H. Hu, Y . Wei, B. Jiang, and Y . Wu, “Ai agent for education: von neumann multi-agent system framework,” arXiv preprint arXiv:2501.00083, 2024

    work page arXiv 2024

  13. [13]

    Adaptive game-based learning in multi-agent educational settings,

    D. Tsatsou, N. Vretos, and P. Daras, “Adaptive game-based learning in multi-agent educational settings,”Journal of Computers in Education, vol. 6, no. 2, pp. 215–239, 2019

    work page 2019

  14. [14]

    Revolutionizing education using person- alized ai: Multi-agent approach,

    V . S. Ganesan and V . Mojes, “Revolutionizing education using person- alized ai: Multi-agent approach,”Available at SSRN 5173671, 2025

    work page 2025

  15. [15]

    Explainable artificial intelligence: a comprehensive review,

    D. Minh, H. X. Wang, Y . F. Li, and T. N. Nguyen, “Explainable artificial intelligence: a comprehensive review,”Artificial Intelligence Review, vol. 55, no. 5, pp. 3503–3568, 2022

    work page 2022

  16. [16]

    Four principles of explainable artificial intelligence,

    P. J. Phillips, P. J. Phillips, C. A. Hahn, P. C. Fontana, A. N. Yates, K. Greene, D. A. Broniatowski, and M. A. Przybocki, “Four principles of explainable artificial intelligence,” 2021

    work page 2021

  17. [17]

    Explainable artificial intelligence in education: A comprehensive review,

    B. A. Chaushi, B. Selimi, A. Chaushi, and M. Apostolova, “Explainable artificial intelligence in education: A comprehensive review,” inWorld conference on explainable artificial intelligence. Springer, 2023, pp. 48–71

    work page 2023

  18. [18]

    Unlocking the black box: an in-depth review on interpretability, explainability, and reliability in deep learning,

    E. S ¸AHiN, N. N. Arslan, and D. ¨Ozdemir, “Unlocking the black box: an in-depth review on interpretability, explainability, and reliability in deep learning,”Neural computing and applications, vol. 37, no. 2, pp. 859–965, 2025

    work page 2025

  19. [19]

    Maestro: Multi- agent educational system for tutoring and recommendation orches- tration,

    R. Saurel, F. Silvestre, J.-B. Raclet, and E. Lescure, “Maestro: Multi- agent educational system for tutoring and recommendation orches- tration,” inEuropean Conference on Technology Enhanced Learning. Springer, 2025, pp. 380–385

    work page 2025

  20. [20]

    Explainable artificial intelligence for education and training,

    K. Fiok, F. V . Farahani, W. Karwowski, and T. Ahram, “Explainable artificial intelligence for education and training,”The Journal of Defense Modeling and Simulation, vol. 19, no. 2, pp. 133–144, 2022

    work page 2022

  21. [21]

    From data to pedagogy: Leverag- ing explainable artificial intelligence to enhance trust, transparency, and effectiveness in intelligent learning systems,

    R. N. Alibayeva and S. U. Allayarova, “From data to pedagogy: Leverag- ing explainable artificial intelligence to enhance trust, transparency, and effectiveness in intelligent learning systems,”Qubahan Techno Journal, vol. 4, no. 3, pp. 51–65, 2025

    work page 2025

  22. [22]

    Multi-agent systems for learning assessment in education: A comprehensive survey,

    R. Zhang, “Multi-agent systems for learning assessment in education: A comprehensive survey,” inProceedings of the 2025 3rd International Conference on Educational Knowledge and Informatization, 2025, pp. 388–391

    work page 2025

  23. [23]

    Requirements en- gineering: A systematic mapping study in agile software development,

    K. Curcio, T. Navarro, A. Malucelli, and S. Reinehr, “Requirements en- gineering: A systematic mapping study in agile software development,” Journal of Systems and Software, vol. 139, pp. 32–50, 2018

    work page 2018

  24. [24]

    Viewpoints: principles, problems and a practical approach to requirements engineering,

    I. Sommerville and P. Sawyer, “Viewpoints: principles, problems and a practical approach to requirements engineering,”Annals of software engineering, vol. 3, no. 1, pp. 101–130, 1997

    work page 1997

  25. [25]

    Requirements engineering: a roadmap,

    B. Nuseibeh and S. Easterbrook, “Requirements engineering: a roadmap,” inProceedings of the Conference on the Future of Software Engineering, 2000, pp. 35–46

    work page 2000

  26. [26]

    Requirements engineering chal- lenges in building ai-based complex systems,

    H. Belani, M. Vukovic, and ˇZ. Car, “Requirements engineering chal- lenges in building ai-based complex systems,” in2019 IEEE 27th international requirements engineering conference workshops (REW). IEEE, 2019, pp. 252–255

    work page 2019

  27. [27]

    Requirements engineering in machine learning projects,

    A. Gjorgjevikj, K. Mishev, L. Antovski, and D. Trajanov, “Requirements engineering in machine learning projects,”IEEE Access, vol. 11, pp. 72 186–72 208, 2023

    work page 2023

  28. [28]

    Requirements engineering (re) in artificial intelligence (ai) systems implementation: The,

    E. Liywalii, “Requirements engineering (re) in artificial intelligence (ai) systems implementation: The,” inArtificial Intelligence Research: 5th Southern African Conference, SACAIR 2024, Bloemfontein, South Africa, December 2–6, 2024, Proceedings. Springer Nature, 2024, p. 423

    work page 2024

  29. [29]

    How to create personas: Three persona creation methodologies with implications for practical employment,

    B. J. Jansen, S.-G. Jung, L. Nielsen, K. W. Guan, and J. Salminen, “How to create personas: Three persona creation methodologies with implications for practical employment,”Pacific Asia Journal of the Association for Information Systems, vol. 14, no. 3, p. 1, 2022

    work page 2022

  30. [30]

    Use of journey maps and personas in software requirements elicitation,

    E. D. Canedo, A. T. S. Calazans, G. R. S. Silva, P. H. T. Costa, and E. T. S. Masson, “Use of journey maps and personas in software requirements elicitation,”International Journal of Software Engineering and Knowledge Engineering, vol. 33, no. 03, pp. 313–342, 2023

    work page 2023